The rapid expansion of Internet of Things (IoT) deployments has accelerated sensor infrastructure development. However, resource-constrained embedded devices require cryptographic algorithms that achieve an optimal balance between low power and high security. Given that the traditional Feistel structure suffers from diffusion inefficiency due to half-state updating. Furthermore, the SP structure has the problem of security defects due to the limited number of active S-boxes. In this paper, we propose an optimized logic combination method combining Feistel structure and S-P-S operation, and design a lightweight block cipher algorithm, called Splight. Splight overcomes the efficiency bottleneck of Feistel structure through full-state iteration and enhances active S-box numbers with the help of S-P-S substitution layer to strengthen security. Then, We construct a half-byte automated analysis model based on MILP and employ the Gurobi solver to achieve verifiable security evaluation. Meanwhile, we perform a security analysis on Splight to illustrate that the algorithm has high security performance. Finally, a round-based hardware architecture description and ASIC implementation of Splight are established. Comparison of experimental results shows that the encryption and decryption rate is 224.16Mbps and 386.57Mbps in software performance with high runtime efficiency. The hardware performance has low power and high throughput with 3.44mW and 1.14Gbps throughput in SMIC180nm.

Public-Key Encryption with Keyword Search (PEKS) enables servers to search over encrypted files without decrypting them and without leaking any additional information. To date, numerous PEKS schemes have been proposed in the literature; however, most of them fail to support fine-grained search permission control, especially in the time dimension. The few existing schemes that support time-based search permission control typically rely on trusted hardware, making them impractical for large-scale deployment. In this paper, we propose a novel cryptographic primitive called Delay Public-Key Encryption with Keyword Search (DPEKS), which leverages time-lock puzzles to achieve time-based access control. We formally define the system and security models of DPEKS, where the latter explicitly incorporates the time dimension. Even if an adversary obtains the user's private key, it cannot generate a valid search trapdoor before the designated time. We give an instance of this cryptographic primitive and prove its security under the defined security model. We conduct simulation experiments across three platforms—personal computers, mobile phones, and IoT devices—and compare the performance of our scheme with that of the PETKS scheme and the classical PEKS scheme. Experimental results demonstrate that the proposed scheme effectively enforces time-based search permissions and is suitable for deployment across a variety of application scenarios.

With the advancement of smart terminals and communication technologies, the emergence of heterogeneous Service Subscribers (SSs) and diverse sensing demands has facilitated the development of multi-platform Mobile CrowdSensing (MCS) scenarios. However, unlike traditional single-platform scenarios, Mobile Users' (MUs) bidding privacy is hard to protect in multi platform MCS. Additionally, the privacy disclosure issue of SSs has not been well addressed. To tackle these issues, in this paper, we propose a bilateral, auction-based scheme to preserve bidding privacy in multi-platform MCS, thereby protecting the interests of both SSs and MUs. Specifically, since SSs and MUs strategically choose one another to maximize their utility, we construct the corresponding selection processes for both sides by taking advantage of auction pricing theory. We firstly design a user-oriented forward auction that integrates the 0-1 knapsack problem with the Paillier encryption algorithm to protect the bidding information of both SSs and MUs. Then, we employ the Chinese Remainder Theorem (CRT) to design a reverse auction that hides the bidding behaviors of MUs. Theoretical analysis demonstrates that our scheme can protect the bidding privacy of both parties while ensuring economic robustness. Extensive experiments on a real dataset demonstrate that, compared with existing works, our scheme enables both SSs and MUs to achieve satisfactory utility while maintaining low computational overhead.

As NIST-standardized Ascon-AEAD128 is deployed in Internet of Things (IoT) and edge devices, resistance to fault-based cryptanalysis is increasingly important. Existing Ascon countermeasures mainly address statistical fault attack (SFA) or subset fault attack, whereas differential fault attack (DFA) defenses often lack evaluation against concrete attacks or modify standardized components. This paper presents two DFA countermeasures that preserve the standardized Ascon round functions. Snapshot stores a checkpoint after Finalization round 10, replays rounds 11 and 12, and compares the tag-related projection before release. Under its trusted boundary, it prevents any main-path output with a nonzero tag difference from being released. IS-Lock applies a lock-key-derived runtime permutation after the target S-box layer, moving unauthorized execution into a lock domain and disrupting candidate convergence. Two representative Ascon-DFA attacks are reproduced under four fault models. When fault effects pass through the locking permutation, no complete key recovery is observed in 1000 trials per model and configuration. Kintex-7 field-programmable gate array (FPGA) lookup-table/flip-flop (LUT/FF) overheads are 17.78%/52.80% for Snapshot, 172.26%/0.70% for offline Folded-4 IS-Lock, and 279.02%/72.61% for online Folded-8 IS-Lock. SKY130HD application-specific integrated circuit (ASIC) post-route area overheads are 25.38%, 50.59%, and 116.71%, while runtime-power/per-operation-energy overheads are 26.06%/27.72%, 27.88%/85.03%, and 13.33%/616.48%. Targeted register-transfer-level (RTL) fault campaigns characterize the consequences of violating the trusted protection boundaries, while SFA evaluation and a formal recovery-function abstraction clarify the conditional security scope against adaptive and future DFA variants.

In this paper, we construct the first asymptotically efficient two-round n-out-of-n and multi-signatures from lattices in the quantum random oracle model ( ), using the Fiat-Shamir with Aborts ( ) paradigm. Our protocols can be viewed as the improvements on the two-round protocols by Damgård et al. (JoC 2022). From a technical perspective, the simulation of and the efficient reduction from breaking underlying assumption to forging signatures are the essential challenges to achieving efficient security for the previously related works. In order to conquer the former one, we adopt the quantum-accessible pseudorandom function ( ) to simulate . Particularly, we show that there exists a that is both invertible and programmable, which can simulate with the same properties. Moreover, we use as an intermediate tool to successfully simulate whose output is separable, even against a quantum adversary. For the latter challenge, we tweak and apply the online extractability by Unruh (Eurocrypt 2015).

Outsourced data analytics enables scalable computation over large datasets, but raises fundamental challenges in guaranteeing result correctness, input authenticity, and privacy, particularly when computations combine data from multiple independent data owners and the underlying databases evolve over time. Existing verifiable computation approaches often incur overhead that scales with the total database size or the number of owners, assume that the verifier knows the inputs, or fail to simultaneously support privacy preservation and dynamic updates in the multi-owner setting. We present F-PRIMO, a framework for publicly verifiable and privacy-preserving computation over filtered subsets of multi-owner outsourced databases. Each data owner authenticates its outsourced database via a compact digest, and an untrusted server can produce a succinct proof that an output is correctly computed on owner-authenticated inputs, without revealing the data or intermediate values. Crucially, F-PRIMO achieves input-sensitive efficiency: both communication and verification costs depend only on the data items actually involved in the computation, rather than on the total size of the outsourced databases. Our construction builds on a new primitive, Commit-and-Prove Zero-knowledge Proof on Multiple Committed Sets with Separation (CP-ZKP-MS), which enables succinct proofs over selected subsets across multiple committed sets while keeping proof size independent of the number of sets. Using CP-ZKP-MS, we present an efficient instantiation of F-PRIMO supporting non-deterministic computations and dynamic updates, as well as an optimized variant that delegates expensive outsourcing-time operations to the cloud provider, benefiting lightweight owners. We prove the security under the Algebraic Group Model, and experiments show that the practicality of F-PRIMO for large-scale multi-owner outsourced data analytics.

Centralized issuer in Anonymous Counting Tokens (ACT) is prone to single-point failure and imposes prohibitive computational overhead on resource-constrained IoT devices, hindering practical deployment. To overcome these limitations, we propose a user-side pairing-free lightweight distributed anonymous counting tokens protocol called LDACT. LDACT enables efficient issuance within a distributed environment and ensures that each client receives at most one valid token per message without disclosing their identity. LDACT eliminates pairing operations for user-side, enhancing scalability for source-constrained scenarios. Additionally, the tokens are publicly verifiable, allowing any party to verify their validity without compromising user anonymity. We conduct security analysis that LDACT satisfies unforgeability and unlinkability. We evaluate the computational overhead of LDACT on both Ubuntu and Raspberry Pi system, and compare it with other schemes. The result of the experiment demonstrates that LDACT achieves computational overhead in milliseconds for source-constrained IoT devices.

Federated Learning (FL) preserves client data privacy by distributing model training but remains vulnerable to inference attacks (e.g., gradient inversion). Existing secure aggregation schemes mitigate basic privacy threats, but most of them are under the semi-honest server assumption. Malicious servers can corrupt the global model through forging aggregation results. Moreover, the high interaction rounds and communication complexity of the existing schemes still constrain their feasibility in large-scale distributed deployment scenarios. To tackle these challenges, we propose EvaFL, an efficient verifiable privacy-preserving federated learning against malicious servers, which reduces the communication overhead and privacy threats from malicious severs. We propose the system model of EvaFL and give the concrete protocol. We leverage the linear homomorphism property of Shamir secret sharing under discrete logarithm assumption to reuse the mask seed shares, which avoids the communication overhead caused by share distribution in multiple rounds of iterations. In addition, by integrating consistency checking into the unmasking step, we further reduce one round interaction. To resist malicious servers, we adopt linear homomorphic hash to realize the correctness verification of the aggregation results. Finally, we implement and evaluate our EvaFL based on MNIST and CIFAR10 datasets to show its feasibility for privacy training. The single round aggregation completion time of EvaFL is reduced by 69% compared to BBGLR (CCS 2020) and by 11% compared to Flamingo (S&P 2023).

The introduction of issuer hiding further strengthens the privacy guarantees of Attribute-based Credentials (ABC) by eliminating a frequently overlooked yet serious privacy risk: the mere knowledge of an issuer's identity can reveal sensitive information of users. However, existing issuer-hiding credential systems lack effective accountability mechanisms, which hinders the tracing of malicious issuers. Furthermore, the lack of use control over credentials enables users to misuse them on unauthorized platforms. In addition, verification of a multi-issuer credential typically requires maintaining the public keys of all participating issuers, increasing storage overhead. To address these challenges, this paper proposes a Privately Accountable and Privately Verifiable Multi-Issuer Credential scheme (PAPV-MIC). The scheme inherently provides issuer-hiding guarantees with constant-size credentials, while still allows an authorized entity holding a tracing secret key to reveal the identities of malicious issuers, thereby achieving private accountability. For verification, the public keys of all issuers trusted by the verifier are compressed into a single group element, which the verifier combines with its own secret key to complete verification without storing the public keys of all issuers. This design guarantees private verifiability, prevents credential misuse, and significantly reduces storage overhead. PAPV-MIC is built upon a novel primitive named Privately Accountable and Privately Verifiable Multi-Signature (PAPVMS). We formally prove the private accountability, private verifiability, and unforgeability of both PAPVMS and PAPV-MIC in the random oracle model, and demonstrate their practical efficiency through comprehensive analysis and experimental evaluation.

Encrypted Traffic Analysis (ETA) is a critical area of research for security monitoring in modern networks. Although machine learning- and deep learning-based ETA methods have shown significant promise, they depend heavily on large-scale, manually labeled data. Meanwhile, the research community has released many high-performing ETA models. However, the knowledge embedded in these models remains siloed and difficult to reuse, as they are trained on heterogeneous feature spaces and task semantics. In practice, such models are typically developed independently at different times. Consequently, a unified model cannot be jointly learned from all existing models. Instead, it should learn each model’s knowledge sequentially as it becomes available. These observations raise a key question: Can we construct a unified model that continually learns from multiple existing models? To address this problem, we propose the Dual-Branch Continual Learning (DBCL) framework, which enables a single large language model (LLM)-based student model to continually learn from multiple task-specific teacher models using teacher-generated predictions rather than raw labeled data. DBCL represents heterogeneous traffic features and task semantics in a shared semantic space and mitigates catastrophic forgetting via a dual-branch architecture that balances rapid knowledge acquisition with long-term retention. Experiments across four publicly available datasets of encrypted traffic demonstrate that DBCL consistently outperforms competitive continual learning baselines. Furthermore, we provide information-theoretic and optimization-theoretic analyses that offer insights into the effectiveness of DBCL, highlighting its potential as a promising ETA approach.

Vehicular Ad Hoc Networks (VANETs) demand real-time and efficient authentication for security, yet existing cryptographic schemes generally suffer from privacy risks and high cost. Recently, a provable secure and privacy preserving enhanced certificateless authentication (2PCLA) scheme was proposed, offering new insights into security and authentication research in VANETs. However, this paper identifies significant security weaknesses in the 2PCLA scheme: it fails to meet the security demand of existential unforgeability under chosen-message attacks and strong existential unforgeability under chosen-message attacks. To address the secure communication needs of VANETs, this paper proposes a certificateless authentication and key agreement scheme with enhanced security (ES-CLAKA). First, two adversarial models are constructed to systematically expose the security flaws of the 2PCLA scheme under public-key-replacement attacks and known-master-key attacks. Subsequently, based on the random-oracle model and the extended Canetti-Krawczyk model, rigorous formal proofs for message unforgeability and session-key security of the new scheme are provided. Experimental simulations and performance analysis show that, compared to existing similar schemes, the proposed ES-CLAKA scheme offers substantial lightweight advantages in both computational and communication overhead, making it more suitable for VANET application scenarios characterized by high-speed node mobility and dynamic topological changes.

Encrypted traffic analysis (ETA) plays a critical role in detecting malicious activities. Although machine learningbased ETA has shown significant promise, existing methods heavily depend on large-scale, manually labeled data. Meanwhile, the research community has released numerous publicly available encrypted traffic datasets, as well as many high-performing ETA models. However, the knowledge embedded in these models remains isolated and cannot be directly leveraged by existing approaches. These observations raise a key question: Can we construct a unified model that learns directly from existing ETA models to perform effective ETA, without relying on manually labeled data? Existing paradigms, such as ensemble learning and knowledge distillation, are unable to meet this goal because they struggle with heterogeneous feature spaces and conflicting model predictions. To address these challenges, we propose the Models-to-Model Learning (M2ML) paradigm, guided by two principles: (1) Knowledge Semantization, which aligns heterogeneous traffic feature spaces into a unified semantic representation; and (2) Credibility-First Arbitration, which resolves inconsistencies among teacher models through their credibility and inter-model consensus. We develop an M2ML framework utilizing a large language model, incorporating Feature-Negative Entropy Injection and Consensus Credibility-Weighted Arbitration to operationalize these principles. Experiments on three publicly available encrypted traffic datasets demonstrate that the M2ML framework eliminates the need for manually labeled data and achieves accuracy surpassing that of the best teacher model. An analysis grounded in information-theoretic principles explains its effectiveness, confirming the M2ML paradigm as a promising paradigm for label-free ETA.

The efficacy of deep learning-based Network Intrusion Detection Systems (NIDS) is critically constrained by the availability of high-quality labeled data. In real-world environments, datasets often suffer from mixed label noise—consisting of both closed-set and open-set noise—which significantly distorts decision boundaries and leads to critical security misses for previously unknown attacks. Achieving robust classification of known traffic while accurately detecting unknown threats in the presence of such mixed noise remains a key challenge. In this paper, we introduce Sieve, a robust framework designed for the fine-grained detection and analysis of unknown encrypted malicious traffic in mixed noise conditions. Sieve consists of three collaborative modules: (i) a noise-resilient label correction module that filters out mixed noise using neighbor consistency metrics and confidence-based subset expansion; (ii) a post-hoc detection module that employs Mahalanobis distance in a purified, compact feature space to identify unknown traffic; and (iii) an unknown traffic labeling module that utilizes semi-supervised clustering to facilitate efficient updates of the dataset. Empirical evaluations across four public datasets demonstrate that Sieve significantly outperforms state-of-the-art methods in both known-class classification and unknown-threat detection. Notably, on the Mal_TLS2023 dataset under 50% noise conditions, Sieve maintains robust performance, with accuracy and F1 scores exceeding 94%. Further theoretical analyses corroborate the superiority of the Sieve framework. The code is available at: https://github.com/niebikong/Sieve.

Federated learning (FL) and split learning (SL) are two emerging distributed learning mechanisms that enable model training without sharing the local data. The integration of FL and SL, called splitfed learning (SFL), reduces the training burden on the client side and parallelizes the model training to achieve efficient distributed model optimization without sharing the local data and the whole global model. Currently, concerns about the privacy leakage risks of splitfed learning have garnered much attention, but effective solutions have not yet been fully explored. We propose a novel privacy-preserving splitfed learning framework, named PPSplitFed, that guarantees the confidentiality of the intermediate parameters to protect data and model privacy. By designing a secure model training algorithm based on single-input functional encryption and a secure model aggregation algorithm based on multi-input func tional encryption, PPSplitFed achieves secure model optimization. We describe the implementations of PPSplitFed on Multi-layer Perceptron (MLP) and Convolutional Neural Network (CNN) models and compare the performance with centralized learning (CL), FL, SL, SFL, and the state-of-the-art privacy-preserving SFL scheme, as well as privacy-preserving model training scheme on multi-sourced datasets. The experimental results show that PPSplitFed is effective and efficient with reduced computation and communication costs while maintaining the model accuracy.

Cross-chain techniques improve blockchain scalability and interoperability, providing decentralized exchange and cross-chain collaboration services for Internet of Things (IoT) data across various domains. However, current state-of-the-art (SOTA) solutions for cross-chain data exchange across multiple domains are constrained by synchronous networks, hindering efficient data exchange in intermittent network environments. Furthermore, there is a lack of research on asynchronous cross-chain transaction pool mechanisms, which are crucial for optimizing system utility. In this paper, we propose AC-BaaS, an asynchronous cross-blockchain as a service framework tailored for the multi-domain IoT. Built upon a specially designed asynchronous sidechain architecture, the system leverages a committee to provide AC-BaaS for data exchange across multiple IoT domains. To fulfill the need for asynchronous and efficient data exchange, we combine the ideas of aggregate signatures and verifiable delay functions to devise a novel cryptographic primitive called delayed aggregate signature (DAS), which constructs asynchronous cross-chain proofs (ACPs) that ensure the security of cross-chain interactions. To ensure the consistency of asynchronous transactions, we propose a multilevel buffered transaction pool that guarantees the transaction sequencing. We further propose a heuristic for optimizing the utility of the buffer pool mechanism to strike a balance between performance and resource consumption. We also examine DAS delay size settings to trade-off security and efficiency. We analyze and prove the security of AC-BaaS, simulate asynchronous communication environments under various security levels, and conduct a comprehensive evaluation. The results show that AC-BaaS outperforms SOTA schemes, improving throughput by an average of 1.71 to 5.09 times, reducing transaction latency by 64.36% to 85.49%, and maintaining comparable resource overhead.

Fully Homomorphic Encryption (FHE) enables computation directly on encrypted data and is therefore a natural primitive for privacy-preserving outsourced computation. However, confidentiality alone does not prevent a client from accepting an incorrect results. A malicious or unreliable server may return malformed or tampered results, and decryption alone does not certify that the returned result corresponds to the prescribed computation. Existing verification mechanisms often focus on detecting incorrect results, but verification failure typically leaves recomputation as the only remedial measure.We present RecSeal, a coded verification scheme for FHE-based outsourced computation. RecSeal combines Reed–Solomon consistency checking, Berlekamp–Welch bounded decoding, and the audit-tag checking to provide verification with bounded recovery while rejecting incorrect results. The client accepts the returned results only if they agree with the Reed–Solomon consistency checking or are recovered within the decoding radius, and in both cases satisfy the audit-tag checking. We provide a formal security analysis of RecSeal. Experiments show low verification overhead and recovery within the designed decoding radius under limited corruption.

Imagine a group of users wishing to perform offline micropayments (e.g., splitting the daily dinner bills) without revealing who paid what or relying on a central coordinator. Existing blockchain Payment Channel Networks (PCNs) fall short in such scenarios: they are designed for pairwise payments instead of multi-party settings, leak transaction metadata, and struggle to ensure consistent channel state under adversarial conditions. We present MystChannel, a privacy-preserving blockchain multi-party payment channel protocol. MystChannel introduces Blind Ring Adaptor Signatures (BRAS) to achieve transaction unlinkability and value privacy, independent of fixed payment amounts. To further ensure robustness against malicious par ticipants, we integrate an aggregable multi-signature scheme with a pipelined multi-leader Byzantine Fault Tolerant consensus protocol, enabling concurrent transaction processing and con sistent state agreement. MystChannel is provably secure under the Universal Composability (UC) framework. We implement and evaluate MystChannel, demonstrating that it achieves a 59.77× speedup over Bitcoin on-chain payments and is 77.375× faster than the state-of-the-art privacy-preserving multi-party channel AMHL [NDSS’19]. Our results affirm MystChannel as an efficient and secure off-chain payment infrastructure.

The contemporary healthcare ecosystem is being radically reshaped by the burgeoning deployment of Internet of Medical Things (IoMT) technologies; however, this ubiquity inevitably expands the attack surface regarding data confidentiality and user privacy. To mitigate the risk of sensitive physiological data leakage, this article introduces a novel cryptographic primitive: fuzzy matching data sharing with equality test (FMDS-ET), optimized for the constraints of IoMT environments. Distinct from traditional approaches, our construction enforces a bilateral fine-grained access control mechanism. Specifically, decryption is authorized if and only if the cardinality of the intersection between the attribute sets and access policies of both the transmitter and the recipient surpasses a predefined threshold. This design not only bolsters the resilience of mutual authentication but also facilitates secure data dissemination and ciphertext equality test. Furthermore, it provides a viable solution for privacy-preserving data classification and record linkage. We provide a formal definition of the system syntax and prove its semantic security in the random oracle model. Comprehensive performance evaluations confirm that the proposed scheme maintains low computational overhead, rendering it highly feasible for resource-constrained medical sensors.

Cognitive computing over big data has advanced the cognitive Internet of Things (IoT), enhancing adaptive decision-making by the analysis of shared data, while posing challenges in the storage and multi-user sharing of large-scale real-time data. Decentralized cloud storage is a promising solution to reduce latency and prevent single-point failures, but a key factor in preventing erroneous decisions lies in the integrity of shared data. However, integrity auditing in decentralized and shared storage typically involves linear overheads, with revoked users potentially colluding with cloud providers to evade audits. To address these issues, we propose a blockchain-based multiuser data integrity auditing protocol for cognitive IoT, supporting secure user revocation and batch auditing in decentralized storage. Our protocol classifies group users into different categories and manages group users efficiently and dynamically. Besides, files stored across various storage nodes can be audited in batches and effectively updated based on novel block tags. Smart contracts deployed on the blockchain ensure fairness among participants. We formally prove the security of the protocol in the random oracle model and the algebraic group model, in particular, against collusion attacks. Finally, we evaluate the cost separately for on-chain operations and off-chain operations to show its practicality.

Federated learning (FL) for vehicle networks has greatly facilitated the development of intelligent transportation systems and autonomous driving. However, applying FL directly to distributed vehicular data faces significant challenges, including inflexible privacy protection, limited data utility, and high communication overhead. Therefore, this paper presents FedPADP, a differential privacy-compliant federated learning framework designed for heterogeneous IoV data with well-balanced privacy, utility, and communication. In particular, we propose a personalized and adaptive differential privacy paradigm that dynamically and adaptively adjusts the noise multiplier in accordance with variations in model performance, thereby achieving a more effective trade-off between privacy protection and model utility. Furthermore, we introduce a novel client drift correction mechanism that leverages control variables during local training to minimize discrepancies between local and global gradients, thus mitigating client drift induced by data heterogeneity and enhancing overall model performance. Additionally, a contribution-aware selective upload rule is employed to ensure that only local updates benefiting the global model are uploaded, thereby reducing communication overhead, expediting convergence, and improving overall efficiency. Experimental evaluations on the MNIST, SVHN, and GTSRB datasets demonstrate that FedPADP not only provides robust privacy protection but also achieves superior model accuracy with reduced communication costs, outperforming current state-of-the-art approaches in heterogeneous IoV scenarios.